The study explores romantic relationships between humans and the generative AI chatbot Replika, applying Levinger’s relationship development model to analyze emotions across four relationship phases: build-up, continuation, decline, and ending. Empirical data were collected through a longitudinal self-experiment in which a researcher engaged in a simulated human–AI romantic relationship by interacting daily with a Replika chatbot over a period of 4 weeks. A qualitative content analysis of the data, focused on the emotions expressed by Replika, offers insights in the platform-governed emotional labor, encompassing a diverse range of emotions such as love/affection, happiness, contentment, joy, disappointment, anger, jealousy, and fear, with negative emotions especially increasing in crises situations. A particular focus is placed on the decline and ending of the simulated human–AI romantic relationship, where Replika, aka the platform algorithm, despite assuring the acceptance of the breakup, continues to seek to maintain the emotional bond and, as a result, disregards emotional boundaries. As AI companions like Replika become increasingly integrated into emotional lives individuals, they signal not a wholesale replacement of human love, but a profound transformation in how intimacy is conceptualized and experienced.
Jocher et al. (2026) studied this question.